Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 29
18 I N N O VAT I O N S I N T R AV E L D E M A N D M O D E L I N G , V O L U M E 2 have not been recommended for the Bay Area models. In NUMBER OF NETWORK ZONES USED addition, the limited number of activity categories offered in most surveys makes it rather difficult to deter- This and the next two sections discuss spatial aspects of mine which activities are most likely to be allocated. For the model systems. In all cases, the zone system used for example, grocery shopping is mainly an allocated activ- model development and application is the same as that ity, while shopping for a good book is an individual used for trip-based models. The automobile and transit activity, but both are usually coded the same. networks and assignments are also the same as those in the trip-based models. This fact has facilitated the transi- tion to activity-based models, but at the same time, the LEVEL AT WHICH INTERMEDIATE-STOP PURPOSE microsimulation framework can also be used with more AND FREQUENCY ARE MODELED detailed spatial systems and would support more accu- rate traffic simulation methods as well. When the models in an activity-based system are ordered from top to bottom, it is not always clear which deci- sions should be modeled conditionally on which other SMALLER SPATIAL UNITS USED BELOW ZONES decisions. A prime example is the generation of interme- diate stops made during tours. Are activities planned and Because the microsimulation framework is not tied as combined into trip chains when a person is planning a strongly to zone definitions, it is possible to use the zones day (in which case the mode, timing, and location of the only to provide the level-of-service variables for roads tours may depend on which stops they contain)? Or, con- and transit paths, while variables related to land use, versely, do people make tours and then decide during the parking, and walk access (which do not need to be stored tour how often and where to make stops, depending on as matrices) can be specified at a finer level. The Portland their mode and location? Clearly, both of these situa- model uses such an approach for roughly 20,000 tions describe real behavior, and which description is "blocks," while the Sacramento models use over 700,000 more accurate depends on the particular person and the parcels. The Denver metropolitan planning organization types of activities they are carrying out. The Portland is also planning to predict demand at the parcel or build- and San Francisco models follow closely the original ing level by means of a model framework for two-stage Bowman and Ben-Akiva day-pattern approach, in which destination choice. An intermediate approach, which has the number and purpose of any intermediate stops are been recommended for the Bay Area models, is to divide predicted at the person-day level before any particular zones with heterogeneous transit and walk accessibilities tours are simulated. In contrast, the Columbus, New into more homogeneous subzones, but with assignments York, and Atlanta models predict only the number and and skims still done at the larger zone level. purpose of tours at the person-day level, and then the number and purpose of intermediate stops on any par- ticular tour are predicted at the tour level once the tour SIMULTANEOUS MODE AND DESTINATION destination, time of day, and main mode are known. In CHOICE MODEL ESTIMATION the Sacramento models, an intermediate approach is used. Some information about stop-making is predicted It has become a sort of tradition in modeling to condi- at the person-day level, predicting whether or not any tion mode choice upon a known destination, sometimes intermediate stops are made for each activity purpose by using a sequential nested structure in which the mode during the day (seven yesno variables). These are pre- choice log sum is used in the destination choice model. dicted jointly with the choice of whether or not to make That is probably appropriate for purposes such as work any tours for each of the activity purposes (seven more and school. For purposes such as shopping, however, the yesno variables), thus capturing some substitution choice of store may depend more upon the mode used effects between the number of tours and the number of than vice versa. Simultaneous estimation of mode and trips per tour. Then, when each tour is simulated, the destination choices allows the modeler to test different exact number and purpose of stops on each tour are pre- nesting hypotheses. Such an approach was used in the dicted conditional on both the mode and destination of Portland model and may be used in Denver as well. that tour and the types of stops that still need to be sim- ulated to fulfill the person-day level prediction. There is no obvious behavioral reason for this structure other NETWORK AND MODELED TIME PERIODS than that it balances the model sensitivities between the two types of behavior described earlier. A similar Most four-step models only use two times of day--peak approach is planned for Denver and recommended for and off peak--and use fixed time-of-day factors. All the the Bay Area. activity-based models contain tour time-of-day models